Demystifying Funding: Reconstructing a Unified Dataset of the UK Funding Lifecycle

Paper · 2026 · William Thorne, Rupert Shepherd and Diana Maynard

Summary

UKRI's Gateway to Research (GtR) records funded UK projects. It does not connect them to the funding calls that produced them, and it says nothing about proposals that panels rejected or postponed. This blocks fairness research on the stages before an award. The authors fill this gap by combining three sources into one PostgreSQL database:

the full GtR data, taken from all of its endpoints;
a snapshot of every UKRI funding opportunity, parsed into a tree of document sections;
panel meeting outcomes and attendance for each research council.

Collecting the panel data was difficult. Councils publish it in inconsistent spreadsheets and PDFs, and some use Tableau dashboards that block exports. To pull award sizes, UKRI contribution share and duration out of opportunity prose, the authors treat the task as closed-domain question answering. Their hierarchical retrieval uses section-based chunks and scores them with a hybrid of BM25 and dense similarity, and it outperforms feeding the model the whole document. The resulting resource lets researchers follow the funding process from call to outcome. Both data and code are publicly available.

Key points

Panel extraction covers all seven councils: 1,450 meetings, 40,509 applications and 11,730 recorded panel appearances. EPSRC, post-2018 AHRC and MRC from 2025 onward are reachable only through export-disabled Tableau dashboards.
The best metadata extraction setup reaches 87.7% overall accuracy, 2.4 points above the full-document baseline. It uses hierarchical chunks, a hybrid weight of α=0.25 and no re-ranker, with dense scoring by all-MiniLM-L6-v2.
Linking is deliberately conservative. About 27.1% of applications are matched to GtR projects, mostly because unfunded proposals have no GtR record. About 29.6% are matched to opportunities through fuzzy title matching.
The dataset inherits GtR's own quality problems, such as orphaned records and uneven historical reporting.
The release is available at github.com/wrmthorne/GtR-Extended.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Institutions

🎓University of SheffieldSheffield ·

United Kingdom113 works · 21,542 citations

Also on the byline: The National Gallery.

Fields, topics and concepts

📚 Shelves 5Computer Science · 16,667 works Information Systems · 1,437 works Decision Sciences · 209 works Information Systems and Management · 41 works Statistics, Probability and Uncertainty · 61 works

Cite

Cite as William Thorne, Rupert Shepherd and Diana Maynard. 2026. Demystifying Funding: Reconstructing a Unified Dataset of the UK Funding Lifecycle. Proceedings of Natural Scientific Language Processing (NSLP) @ LREC 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/2dqfnwsvr8as

@inproceedings{thorne2026demystifying,
  title = {{Demystifying Funding: Reconstructing a Unified Dataset of the UK Funding Lifecycle}},
  author = {William Thorne, Rupert Shepherd and Diana Maynard},
  booktitle = {Proceedings of Natural Scientific Language Processing (NSLP) @ LREC 2026},
  year = {2026},
  doi = {10.63317/2dqfnwsvr8as},
  url = {https://doi.org/10.63317/2dqfnwsvr8as}
}

References

2 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

Meetings by Council Over Time The distribution of meetings for each of UKRI’s funding councils over time can be seen in Figure 2. find
Domhnall Carlin and Austen Rainer. 2025. Tracking research software outputs in the UK. ArXiv:2507.22871 [cs]. — [arXiv:2507.22871](https://arxiv.org/abs/2507.22871) find

Updated 1 November 2026

This page in Markdown → https://md.research.elamur.ai/free/en/papers/thorne-2026-demystifying-funding-reconstructing-unified-dataset/